🤖 Python and Ship

Install and manage Python versions with pyenv or native python3.

Without the right version, nothing works. Claude Code needs a consistent Python setup.

pyenv install, pyenv local, python3 --version, PATH.

Virtual environments isolate dependencies by project.

Prevents conflicts and ensures reproducibility.

python3 -m venv, uv venv, activate, .venv in .gitignore.

Install packages with pip or uv (a fast alternative to pip).

Claude Code generates requirements.txt; you need to know how to install dependencies and pin versions.

pip install, pip freeze, requirements.txt, uv pip install.

Write simple Python scripts with help from Claude Code.

Automating repetitive tasks is the first step toward "shipping."

argparse, pathlib, sys.argv, if __name__.

Use Claude Code to identify and fix bugs quickly.

Debugging takes 50% of development time; AI reduces that dramatically.

Paste traceback, ask for a fix, breakpoints, pdb.

Add type hints with Claude’s help for more robust code.

Type hints improve autocomplete and reduce runtime bugs.

str, int, list[str], Optional, TypedDict, mypy.

Extract data from web pages with requests and BeautifulSoup.

Automation starts with data; scraping is the gateway.

requests.get, BeautifulSoup, selectors, rate limiting.

Process hundreds of files with pathlib and glob.

Manual tasks that take hours become a 20-line script.

pathlib.Path, glob, shutil, bulk rename.

Read, transform, and export data between CSV, JSON, and other formats.

Every data pipeline needs to convert formats.

csv.DictReader, json.load/dump, basic pandas.

Schedule scripts to run automatically with cron/systemd.

Real automation = running without human intervention.

crontab -e, cron syntax, logging, systemd timers.

Consume REST APIs from services like GitHub, Slack, etc.

Powerful scripts connect services to each other.

requests, headers, auth tokens, pagination.

try/except, logging, retries, and graceful degradation.

A production script can't die silently.

try/except, logging module, tenacity, exit codes.

Create your first API with FastAPI in minutes.

FastAPI is the fastest Python framework for modern APIs.

FastAPI(), @app.get, uvicorn, async.

Define data schemas with Pydantic BaseModel.

Automatic validation eliminates data bugs at the point of entry.

BaseModel, Field, validators, response_model.

Implement the four basic operations of any API.

CRUD is the universal pattern; every API starts here.

@app.post/get/put/delete, status codes, path params.

Add CORS, rate limiting, and basic authentication.

An API without auth is an invitation to disaster.

CORSMiddleware, Depends, OAuth2, API keys.

FastAPI automatically generates Swagger UI and ReDoc.

Living documentation that makes debugging and onboarding easier.

/docs, /redoc, OpenAPI schema, tags.

Run locally with uvicorn and deploy remotely with gunicorn.

An API that only runs locally isn’t a product.

uvicorn, gunicorn, reverse proxy, HTTPS.

Write and run tests with pytest.

Tests are the safety net that allows for fast iteration.

test_ prefix, assert, pytest -v, markers.

Fixtures set up state; conftest.py shares it across files.

Prevents duplication and keeps tests clean.

@pytest.fixture, scope, conftest.py, parametrize.

Simulate external dependencies with unittest.mock.

Test without depending on a database, API, or filesystem.

Mock, patch, MagicMock, side_effect.

Measure which lines of code are exercised by the tests.

You don't know what isn't tested until you measure it.

pytest-cov, --cov=src, coverage html, branch coverage.

Write the test before the code, with Claude generating the implementation.

TDD + AI is a superpower: you define what, AI handles how.

Red-green-refactor, test-first, specification by example.

Run tests automatically on every push/PR.

CI ensures no one (not even AI) breaks what works.

GitHub Actions, pytest step, coverage badge, fail fast.

Package your Python app in a Docker container.

"Works on my machine" becomes "works everywhere".

Dockerfile, docker build, docker run, multi-stage.

Orchestrate multiple containers (app + DB + redis) with a YAML file.

Real apps have more than one service.

docker-compose.yml, services, volumes, networks.

Deploy with one command on platforms like Railway or Fly.io.

Cloud PaaS is the fastest way to get your app online.

railway deploy, fly launch, Procfile, buildpacks.

Manage .env, secrets, and configs by environment (dev/staging/prod).

Leaked secrets = disaster; wrong config = downtime.

.env, python-dotenv, pydantic-settings, 12-factor.

Structured logging, health checks, and basic alerts.

Without monitoring, you only discover problems when the user complains.

structlog, /health endpoint, uptime monitoring.

Horizontal vs. vertical scaling, caching, and when to consider them.

Knowing when (and when NOT) to scale avoids overengineering.

Replicas, load balancer, Redis cache, premature optimization.

Build the minimum viable solution that delivers real value.

Perfectionism kills projects; MVPs launch projects.

MVP, scope cutting, “good enough,” time-boxed.

Deciding what NOT to do is just as important as deciding what to do.

With AI, you can build quickly; deciding WHAT becomes the bottleneck.

Must/should/could/won't, impact vs effort, backlog grooming.

Pre-release checklist: tests, docs, monitoring, rollback.

Launches without a checklist are launches with surprises.

Pre-launch, smoke tests, rollback plan, feature flags.

Collect real feedback and turn it into improvements.

A product without feedback is guesswork.

User feedback, analytics, bug reports, feature requests.

Short build-measure-learn cycles after launch.

The launch isn't the end; it's the beginning.

Sprints, changelogs, semver, continuous deployment.

Paths after mastering Python + Claude Code: SaaS, open source, consulting.

Knowing the destination helps you choose the next step.

SaaS, open source, freelance, portfolio, community.